Case Study: AgTech

Teralytic – Soil Management using IoT

Agriculture is the world’s largest industry and the least digitized.  Over the past 100 years, the world’s soil health has been degrading at an alarming rate, farmers continue to use excessive amounts of irrigation and fertilizer, and it results in tens of billions of dollars in waste and hundreds of billions of dollars in environmental damage and cleanup.  

Teralytic developed the world’s first wireless NPK (nitrogen, phosphorous, potassium) sensor to solve this problem. Multiple large firms have tried and failed at solving this problem over the years, yet demand is still here.  Not only does the sensor monitor NPK, it monitors moisture, respiration, pH, ammonium (among others).

Ultimately, the most valuable asset Teralytic has is the data set their sensors are generating.  For the first time ever, farmers, agronomists, and environmental scientists can observe how the nutrients, moisture, and other sensor readings are related and respond under a variety of conditions and inputs.

Teralytic came to us to architect and build the IoT infrastructure to handle their data at a global scale across millions of devices. Merely collecting and providing the data to consumers was not nearly enough.  They also needed us to build the analytics, machine learning, and data visualization of the sensor data in context with third party data sources such as weather station data, SSURGO soil database and lab reports.  Using ML and the APIs we built, they can also generate global soil nutrient models, as well as individual field level models. Other applications of ML that we are currently implementing involve sensor calibration, anomaly detection and predictive analytics of future sensor readings.

The Teralytic platform current runs a combination of AWS Services from EKS, ECS, Kinesis, API Gateway, Batch, Lambda, RDS.  Services were written in Go, Python (Keras and TensorFlow). The front end was developed using ReactJS.